DEGREE PROJECT IN TECHNOLOGY, FIRST CYCLE, 15 CREDITS
STOCKHOLM, SWEDEN 2020
Parameter
Guidelines for Electric Vehicle Route Planning
Joakim Fridlund and Oliver Wilén
KTH ROYAL INSTITUTE OF TECHNOLOGY
Abstract
There is an urgent need to migrate the vehicle industry from conventional combustion vehicles to electric vehicles due to pressing climate changes caused by the fossil fuel industry. The general public seem to have a prejudice against electric vehicles due to their limited range and the extra planning that may be required from the user. The market for electric vehicles is much more limited than for conventional vehicles, partially because it is a much younger industry. Buying an electric vehicle is also a bigger change than just buying a new car, one has to plan and manage the limited range in a new way. Unfortunately, it is more complicated to create a route planner for electric vehicles than for conventional vehicles and the market for such planners is limited. The complication is because an optimal route is better calculated by lowest energy consumption, rather than the shortest path. This requires more parameters in the routing algorithm to accurately calculate the energy consumption for individual vehicles.
The problem attended to in this thesis is that no clear guidelines exist about which parameter affect the energy consumption in an electric vehicle and to what degree. The purpose of this thesis is to provide guidelines that can show which of nine chosen parameters to implement in an electric vehicle route planner. The parameters chosen in this thesis are already implemented in Simulation of Urban Mobility, a road traffic simulator.
The simulator is used in this thesis to simulate electric vehicles with different parameter values and analyse the impact they have on the energy consumption when the values are incremented.
The thesis shows that although some parameters have a relatively large impact on the energy consumption, it is hard to approximate the correct values for them, and therefore not worth implementing.
Keywords
Electric vehicle, Guidelines, Simulation of urban mobility, Electric route planner implementation
Sammanfattning
Det finns ett brådskande behov att migrera bilindustrin från fossildrivna bilar till eldrivna bilar på grund av den rådande klimatpåverkan av fossila bränslen. Allmänheten verkar ha fördomar mot elbilar på grund av deras begränsade räckvidd och den ytterligare planering som krävs av en elbilsanvändare. Marknaden för elbilar är mer begränsad än marknaden för fossildrivna bilar. Delvis för att elbilsmarknaden är en mycket yngre industri men också för att köpa elbil är en större förändring än att köpa en vanlig bil. En elbilsförare måste använda bilen på ett annorlunda sätt på grund av den kortare räckvidden. Dessvärre så är det mer komplicerat att skapa en ruttplanerare för elbilar än för fossildrivna bilar, och marknaden för sådana ruttplanerare är begränsad. Problemet är att en optimal rutt för en elbil är beräknas mer effektivt med hjälp av lägsta energikonsumtionen istället för den kortaste vägen. Detta kräver mer parametrar i algoritmen för ruttplanering för att effektivt beräkna energikonsumtionen för individuella fordon.
Problemet som hanteras i denna rapport är att det inte finns några tydliga
riktlinjer om vilka parametrar som
har störst påverkan på energikonsumtionen i en elbil. Syftet med denna rapport är att förse riktlinjer som visar vilka av nio valda parametrarna som är värda att implementera i en ruttplanerare för elbilar. Parametrarna som valdes är implementerade i trafiksimulatorn Simulation of Urban Mobility. Trafiksimulatorn används för att simulera elbilar och analysera förändringen i energikonsumtionen när parametervärdena stegvis ökas.
Rapporten visar att även om vissa parametrar har en relativt stor påverkan på energikonsumtionen så är det svårt att uppskatta de korrekta värdena för dem. Dessa parametrar är därför inte värda att implementera.
Nyckelord
Elbil, Riktlinjer, Simulation of Urban Mobility, Ruttplanerare elbil
Acknowledgements
Thanks to our supervisor from KTH, Mira Kajko-Mattsson and our examiner Thomas Sjöland. A big thank you to YABS, the consultant firm who helped us weekly with feedback, and enthusiasm, for our work.
Authors
Joakim Fridlund: jfridl@kth.se Oliver Wilén: owilen@kth.se
School of Electrical Engineering and Computer Science, KTH Royal Institute of Technology
Place for Project
YABS - Stockholm, Sweden
Examiner
Thomas Sjöland Stockholm, Sweden
KTH Royal Institute of Technology
Supervisor
Mira Kajko-Mattsson Stockholm, Sweden
KTH Royal Institute of Technology
List of Figures
2.1 Weighted graph . . . 5
3.1 Research Strategy . . . 9
3.2 Research Phases . . . 13
3.3 Road network . . . 15
4.1 Simulation versus real world by destination . . . 23
List of Tables
3.1 Statistics of the random routes . . . 154.1 The parameters that are used in the guidelines . . . 19
4.2 Starting parameter values. Target model: Tesla Model S 70D . . . . 22
5.1 Slopes and intercept . . . 27
5.2 Difference in energy consumption over parameter range . . . 27
5.3 Impact on energy consumption relative to mean energy difference (470.15Wh) . . . 29
5.4 Parameter by difficulty to acquire . . . 31
5.5 Guidelines for implementing electric vehicle parameters in a routing application . . . 33
Contents
1 Introduction 1
1.1 Problem . . . 2
1.2 Purpose . . . 2
1.3 Goal . . . 2
1.4 Research Method . . . 3
1.5 Commissioned Work . . . 3
1.6 Target Audience . . . 3
1.7 Scope and Limitations . . . 3
1.8 Benefits and Sustainability . . . 4
1.9 Thesis Outline . . . 4
2 Theoretical Background 5 2.1 Introduction to Routing . . . 5
2.2 Routing for Conventional Vehicles . . . 5
2.3 Routing for Electric Vehicles . . . 6
2.4 Simulation of Urban Mobility . . . 6
2.5 Simulation Parameters . . . 7
2.6 Related Work . . . 7
3 Method 9 3.1 Overview . . . 9
3.2 Research Method . . . 10
3.3 Research Instruments . . . 10
3.4 Research Phases . . . 12
3.5 Validity Criteria . . . 16
3.6 Ethics & Sustainability . . . 16
4 Research Process 19 4.1 Parameter Definitions . . . 19
4.2 Simulation Results . . . 22
5 Result, Analysis & Discussion 27 5.1 Results . . . 27
5.2 Analysis . . . 28
5.3 Discussion . . . 31
6 Conclusions and Future Work 35 6.1 Conclusion . . . 35
6.2 Future Work . . . 35
References 37
1 Introduction
There is an urgent need to migrate the vehicle industry from conventional combustion engine vehicles to electrical vehicles. The necessity to shift towards electric vehicles is largely motivated by the large dependency on fossil fuels, which has negative effects on air pollution and climate change, as well as the diminishing supply of fossil fuels steadily increasing the overall cost to operate a conventional vehicle [1]. The general public appears to assume that it is problematic to own an electric vehicle, if you plan on driving long trips or live in less urbanised areas [2]. As long as the public opinion is unchanged, the migration towards electric vehicles will be hampered.
When people first think about getting an electric vehicle, they often experience drivers range anxiety. Drivers range anxiety is the fear of not having enough range left in the vehicle to the next charging station. One of the main factors that lead people to have drivers range anxiety appear to be originating from the lack of range that electric vehicles had at the beginning of the 21st century [3]. According to [3], 70% of people asked said they would not buy an electric vehicle until it reached 300 miles (482km) range. This study, and other studies of similar age, are still used in papers written in recent years as references to driver’s anxiety [4, 5]. Today, Tesla Model S (Long range) has a 348 mile (610km) range which far exceeds the range desired and thus should, in theory, decrease the anxiety [6].
There are several other parts of driving an electric vehicle that can cause anxiety, and the inexperienced driver appears to have more anxiety than an experienced driver [7, 8]. If a driver is inexperienced and is discouraged from driving long trips because they are not familiar with the road or are not sure how the charging stations are spread across the route, they might not consider using electric vehicles at all. One way of decreasing the anxiety would be to have a reliable tool that plans the trip for the driver and ensures the driver a pleasant journey.
Planning a trip with an electric vehicle is not a trivial task. Variables such as air drag and vehicle mass both affect the energy an electrical vehicle uses per kilometre. Since trip planning is not easy, creating an application that can plan routes for electric vehicles effectively is not a trivial task either. When using Google Maps or other traditional maps for a trip, only the fastest route is calculated. For example, the current state of charge can not be used as a parameter since it depends on what state the driver’s battery is in at a specific moment.
This, and more parameters, need to be accounted for when making an application specifically designed for electric vehicle trips [9]. This will inevitably make an
algorithm that calculates the route more complex and the application harder to develop.
There exist no guidelines that describe what parameters matters and in what proportions when calculating a route for electric vehicles. Most of the research about route planning has been with focus on energy optimisation only with fixed parameters, such as [10–12]. Energy optimisation is the most important thing when it comes to calculating a route with electric vehicles and what parameters used when optimising is a big factor. If an application implements too many parameters that do not affect the end route or energy consumption, the computational cost will be too high without gaining anything in the result. On the other hand, if for example only constant speed is used when calculating energy consumption, the computational cost will be low but the calculation will output a poor result compared to the reality.
1.1 Problem
When thinking of getting an electric vehicle, or trying one for the first time, most people have a fear of not having enough range left, or running out of battery, before arriving at a charging station [3]. Neither the third-party market of route planning services for electric vehicles or the charging infrastructure is as well established as the one for conventional vehicles. There are few web application or stand-alone applications that take in to account all of the variables needed to plan a trip with an electric vehicle. This creates a divided market as different electric vehicles have different prerequisites. Some vehicles might have a routing planner built in their information or entertainment systems and some might not.
The problem is that no guidelines for electric vehicle route planners exist today, so every developer needs to research and decide for themselves on how to prioritise different parameters in their algorithm.
1.2 Purpose
The purpose of this thesis is to create guidelines for which parameters to implement when creating a route planner for electric vehicles.
1.3 Goal
The goal of this thesis is to enable researchers to create a framework with these guidelines and extend it with more parameters, further research the guidelines
developed in this thesis and include recommendations for routing algorithms.
More clear guidelines will hopefully increase the number of quality route planners for electric vehicles.
1.4 Research Method
The study will use a qualitative comparative research method. To produce data for the study, the electric vehicles will be simulated. The data that is produced is quantitative, but will be analysed in a qualitative manner.
The comparison model used to compare the parameters will be based on two criteria; The relative impact on the energy consumption, and the availability of the parameter value.
1.5 Commissioned Work
This thesis project is done in conjunction with commissioned work from YABS (Young Aces By Sylog). YABS introduced us to the real work problem a colleague experienced with planning a trip with an electric vehicle and helped us with the project from start to finish. YABS is a consulting firm that provides testing, integration of software and IT solutions for their customers. They operate in many different areas such as telecommunication, the defence sector, medicinal technology, vehicle industry and more.
1.6 Target Audience
The target audience of this thesis are engineers, developers and researchers interested in the electric vehicle industry.
1.7 Scope and Limitations
Because of time and funding limitations, the testing will be done by simulation instead of real tests. The complexity of the study is therefore limited to the capabilities and limitations of the simulation software. The study will only consider third party routing for personal vehicles.
Due to constraints in time, the study will produce guidelines for a limited number of parameters. Change in weather and other pure physical variables will not be considered in this thesis.
1.8 Benefits and Sustainability
The benefits of this report is a broader understanding of the state of the electric vehicle market and the common issues that an electric vehicle driver might face regularly.
By decreasing the gap between conventional and electric vehicles, the need for fossil fuel can hopefully decrease and create a more sustainable vehicle industry.
1.9 Thesis Outline
• Section 2: This chapter includes extended background on vehicle routing, vehicle simulation and vehicle parameters.
• Section 3: This chapter describes the research strategy of the study.
• Section 4: This chapter describes presents the data that was collected that influenced the result of the study.
• Section 5: This is the chapter that provides the result, the answers and hard facts that were found in the research. This chapter also includes the analysis and discussion about the presented results.
• Section 6: This chapter summarises the result of the study while also providing examples for future work in the area.
2 Theoretical Background
This chapter provides an introduction to how routing works today and the simulations performed in this study. In Section 2.1 there is a short introduction to routing. In Section 2.2 and 2.3, routing for conventional vehicles and electric vehicles are explained. In Section 2.4 provides information about the tool that is used for simulations in this paper. Section 2.5 describes why. Section 2.6 describes related work.
2.1 Introduction to Routing
To digitally route a vehicle, the real world needs to be represented as a map.
To represent a map, weighted graphs are used, see Figure 2.1. A city, or any destination, can be represented as a node and each road or distance connecting two destinations is considered an edge. The edges are assigned a weight, this represents the cost of going from the first to the second node connecting the edge.
The cost can be either a distance, actual fuel cost or something else depending on the situation.
Figure 2.1: Weighted graph
2.2 Routing for Conventional Vehicles
For conventional vehicles, the routing and cost calculations are rather simple.
The cost can be represented as C = length of road
speed limit , the time it takes to traverse the path while following the speed limit. An extra variable that can affect the cost is potential congestion in the traffic, in which the speed will be lower than the speed limit.
The shortest path between two distant cities is then optimised over all available edges and nodes with a shortest path-algorithm, usually a variant Dijkstra’s Algorithm due to its relatively low time complexity [13]. Conventional vehicles do not need a much more complex algorithm to calculate the route. Internal combustion engines have a close to fixed fuel consumption and have long enough range that it is near impossible to run out of fuel without finding a gas station.
This is what makes the routing very simple for conventional vehicles.
2.3 Routing for Electric Vehicles
For electric vehicles, the optimal path may not always be the shortest. Electric vehicles depend very much on how much energy is used for a route due to their limited range. Therefore, a routing algorithm will need to calculate the cost as energy consumption rather than just the distance and velocity.
To calculate how much energy a vehicle needs to use to propel it forwards also differs from vehicle to vehicle. Everything, from the weight - to the physical properties of the car - to the efficiency of the engine, play a part. Another reason why electric vehicles require a more complex routing algorithm is due to special properties of the engine. Electric motors can use something called multi- directional power flow. This means it can reverse the power flow, allowing it to not only output energy but also absorb it [14]. Also known as regenerative breaking in vehicles, this feature makes it possible for an electric vehicle to gain energy while driving on a road with a downhill slope or when breaking in general.
Regenerative braking can increase cruising range by about 20% in some settings and even further in more hilly areas [15]. Due to this feature, a road that takes more time to traverse might be more suitable for the electric vehicle if it can gain energy by taking that path. This allows for the introduction of negative weights on the road network. Negative weight could then be interpreted as the vehicle is gaining more energy than it uses while traversing it.
2.4 Simulation of Urban Mobility
Simulation of Urban Mobility (SUMO) is an open-source road traffic simulator developed by the Institute of Transportation Systems at the German Aerospace Center [16].
SUMO is an application for microscopic, space-continuous road traffic simulation.
Within the SUMO package, there are numerous tools for simulating multiple
traffic scenarios and vehicles. SUMO provides simulation of electric vehicles with methods to monitor energy consumption.
There are limitations of what parameters that can be adjusted by a user.
Environmental variables and some vehicular variables such as the discharge curve for an electric vehicle battery are not possible to alter without modifying the source code. It would, however, be possible to modify the source code to further customise the behaviour of the simulation.
In a 2018 study [17] the accuracy of electric vehicle simulation in SUMO was questioned. The findings of the study showed that SUMO had a 23-29% difference in energy consumption compared to a test with a real electric vehicle. The study proposed some improvements that can be implemented in SUMO that will improve the simulation accuracy significantly. These are, however, not documented by SUMO as implemented changes.
2.5 Simulation Parameters
When simulating a vehicle of any kind, there are several parameters that need to be considered. This is no different for electric vehicles. To gain an accurate simulation, the parameters need to mimic the characteristics of the real physical parameters that are present in an electric vehicle. However, there is endless complexity in the real world scenario, so simulation programs need to simplify the vehicle to a smaller amount of parameters that contribute the most to the behaviour in a real scenario.
In SUMO, there are physical calculations and several parameters calculating the energy consumption. In this theses, we will focus on nine of them. The parameters chosen are presented in Section 5.1.
2.6 Related Work
Since the rise in popularity of electric vehicles, some implementations of routing programs have been developed. But no guidelines on how to develop such an application has been presented. However, some studies have researched similar areas to our study. The article Electric vehicle modelling and energy- efficient routing using particle swarm optimisation written by Abousleiman R and Rawashdeh O in 2016 [11] is a study on how different routing algorithms predict the energy consumption in a route. The results of the different algorithms were tested by a real electric vehicle. The study concludes that the particle swarm optimisation algorithm is the most effective for larger road in 2019 networks.
The article Development of an Energy-Efficient Routing Algorithm for Electric Vehicles by Garcia et al. [12], where the particle swarm optimisation algorithm is tested in SUMO. The study came to similar results as the 2016 study by Abousleiman R and Rawashdeh O.
3 Method
This chapter presents the research method used in the study. Section 3.1 is a summary of the research strategy used in this study. Section 3.2 describes the research method of the study. Section 3.3 presents the research instruments used in the study. Section 3.4 describes the phases and timeline of the study. Section 3.5 describes the validity criteria that were used to validate the results of the study.
Section 3.6 Describes the ethical criteria that were followed in this study.
Figure 3.1: Research Strategy
3.1 Overview
The research strategy for this thesis project, illustrated in Figure 3.1, was chosen to give flexibility in the literature study phase. This was in part due to time constraints for the project, but also to give us more flexibility if a problem arose with the research method. Therefore, the research strategy that we chose will allow the Literature Study and Preparation Work phases to continue during the other phases. The chosen research method for the study is a qualitative comparative method where the data processed is largely statistical and the analysis is context-based. The method aims to compare the importance of parameters in electric vehicle route planning. The process of the study is represented in five phases, which are explained in detail in 3.4. The research
instruments of the study were chosen to produce valid data efficiently. The testing was made in SUMO and the results were validated against a real electric vehicle.
The validity of the study was analysed by observing the results with qualitative validity criteria. The last step of the research strategy was to analyse the research process from an ethical standpoint.
3.2 Research Method
Qualitative and quantitative are the two major research approaches. The quantitative approach relies heavily on statistics and figures. It focuses on statistical, mathematical and computational techniques. According to [18], quantitative research is most effective on large sample sizes where a generalisation of a large population is desired. A disadvantage of quantitative research is that the context is usually treated as noise. And as such, quantitative research trades the context for generalisation over a population.
The qualitative research approach focuses on investigating things without comparing them in a statistical manner. It is effective when a deeper study on a particular subject. Qualitative research is well suited for exploratory research when the subject is not well researched yet [18]. A disadvantage of qualitative research is that it is difficult to generalise to a larger population.
In this thesis, a qualitative research method will be used since our comparison model requires a qualitative analysis of the data. We will use a qualitative, comparative analysis method where each parameter will be evaluated using our comparison model. We do not use a quantitative method because we do not have such large statistical data set it is required.
3.3 Research Instruments
This section describes the research instruments used in this thesis. Section 3.3.1 describes the citation database Scopus. Section 3.3.2 describes two important papers that influenced the study. Section 3.3.3 describes Simulation of Urban Mobility, the simulation program used in the study. Section 3.3.4 describes OpernStreetMap, the online map tool used to import road networks into the simulations. Section 3.3.5 describes the Tesla model S electric vehicle used to verify the accuracy in the simulations. Section 3.3.6 describes two valuable Python modules that were important in the simulation script.
3.3.1 Scopus
Scopus is a source-neutral abstract and citation database curated by independent subject matter experts [19]. We used Scopus to find most of the references and papers that we have used in our study. Scopus was particularly useful in the Literature Study phase when we needed to find papers in the general scope of our study.
3.3.2 Research papers
We based our choice of research instruments, and partly our research methods on previous studies that researched similar areas to our scope. Among these papers, two studies influenced our research more than the others. One written by Abousleiman R and Rawashdeh O from 2016 [11], where a model characterising the energy consumption of an electric vehicle is presented and applied with the particle swarm optimisation algorithm. And the other by Garcia A et al. which was written in 2019 [12], where the performance of particle swarm optimisation for electric vehicle route planning is tested using Simulation of Urban Mobility.
3.3.3 Simulation of Urban Mobility
We used Simulation of Urban Mobility (SUMO) to simulate how different parameters affect the energy efficiency of a planned route for an electric vehicle [20]. SUMO also includes tools for generating random trips and generating road networks. We used both of these tools in our data collection to generate routes to test the electric vehicles in realistic road scenarios.
3.3.4 OpenStreetMap
OpenStreetMap is a web-based map application that is collaboratively developed by a community of cartographers[21]. In our study, we used OpenStreetMap to acquire road network data for our simulations in SUMO.
3.3.5 Tesla Model S
To verify SUMO’s efficiency, we drove a Tesla Model S 70D. We drove the car on three routes and recreated the same routes and modelled the vehicle as close as possible in SUMO. The energy consumption for the real car and the results from SUMO were compared.
3.3.6 Select Python modules
Python was the chosen programming language and several modules were used.
The list provided is of the most important modules as modules such as Numpy, which is a powerful tool, still can be omitted with the same result.
• xml.etree.ElementTree was used to parse the SUMO-generated XML files and do operations on them.
• Subprocess is a module to run system processes in python. This was used to run SUMO simulations in iterations from within python.
3.4 Research Phases
Our study was divided into five distinct research phases. As illustrated by Figure 3.2, these are Literature Study, Preparation Work, Model Development, Data Collection, and Comparison & Analysis. The Literature Study and Preliminary work are both continuous work throughout the research and are described in Section 3.4.1 and 3.4.2 respectively. The Development of A Comparison Model phase was completed before the Data Collection phase to have clear criteria of what to extract from the data. These sections are presented in Sections 3.4.3 and 3.4.4. The Comparison & Analysis phase is described in Section 3.4.5.
3.4.1 Literature Study
At the beginning of our work, we had limited knowledge about electric vehicle route planners. The goal of this initial study was to gain enough knowledge about the state of electric vehicles and electric vehicle route planners so that the problem of this thesis could be more defined and specific. To achieve this, we would need to gain an understanding of the state of the electric vehicle industry.
We began this phase by searching for relevant scientific papers about electric vehicle route planning and compiling them to a list of possible sources. We used KTHs library searching tool Primo and Scopus as well as IEEE explore and Researchgate to find these scientific papers [19, 22, 23]. The papers were then read by both of us, and depending on their relevancy and quality, were kept as possible sources. Among these papers, a handful of particular importance were found. Two short papers about particle swarm optimisation with electric vehicle route planning. One written by Abousleiman R and Rawashdeh O from 2016 [11], and the other by Garcia A, Tria L and Talampa M from 2019 [12]. These two studies would become important aids in our research, and would inspire the research instruments that we used in this study.
Figure 3.2: Research Phases
3.4.2 Preparation Work
After the initial study, we had gained enough knowledge about the state of electric vehicle route planners and had a more clear understanding of how we would research the subject area. Since the study by Garcia A, Tria L and Talampa M from 2019 used Simulation of Urban Mobility (SUMO) as their means of testing the energy efficiency of a route for an electric vehicle, we felt that SUMO would be an adequate research instrument [12].
We learned how to use SUMO by reading the documentation on their website, and by modelling simple simulations and assessing the output data of the application.
While we were learning and researching SUMO, we came to the conclusion that changing the source code would be too much effort. Therefore, we decided to
focus our attention on the parameters that already were implemented and easily accessible.
We also found a paper by the end of the phase that described flaws in SUMO’s simulation accuracy for electric vehicles [17]. The paper showed that SUMO can have a 23-28% error on the energy consumption of an electric vehicle. Since there was limited time for the project, we decided to continue with SUMO and to validate the simulation results against a real electric vehicle. After this phase, we had a better understanding of the scope of the study based on the advantages and limitations of the research instruments.
3.4.3 Development of Comparison Model
The Comparison Model consists of two comparison criteria; (1) energy consumption and (2) ease of access. The first criteria, (1), was evaluated by simulating cars in SUMO and comparing the difference of the parameters in energy consumption over a range of values commonly found in personal vehicles.
The second criteria, (2), was evaluated by how easily accessible the parameter value is to the public based on previous research. These two criteria together would give a guideline on which parameters for an electric vehicle should be considered first when creating an electric vehicle routing application.
3.4.4 Data Collection
The comparison model needs the energy consumption of several vehicles to compare to each other. That data was generated from SUMO by simulating electric vehicles driving in a virtual world. In SUMO, the world is represented by edges and nodes connected in a network. The edges and nodes represent roads and destinations respectively. SUMO provides a python script called osmWebWizard.py, which serves as a tool to convert road world map from OpenStreetMap, an open-source world map, to a graph with edges and nodes.
The edges have different properties to represent the cost of traversing it. The cost is a combination of several parameters with the primary being length, speed limit and elevation.
The simulation was run on an area in Stockholm shown in Figure 3.3a. Which area to use was decided semi-arbitrarily. The map should not be large enough that the files became too large for the computers to process, and it should be in the Stockholm area. The chosen area can be seen in 3.3a. TheosmWebWizard.py converts the map from OpenStreetMap to a graph with edges and nodes. The interpretation can be seen in 3.3b and is stored in XML-files.
(a) Generated map from OpenStreetMap
(b) SUMO interpretation
Figure 3.3: Road network
Table 3.1: Statistics of the random routes Description Length (km)
Shortest route 2.99
Longest route 13.51
Average distance 6.53
Median distance 6.01
To decrease the bias of one particular route, it was decided to simulate several routes. For this, another tool calledrandomTrips.py (also provided by SUMO) was used to generate multiple trips after the network was generated. These trips were then used to create routes with an electric vehicle attached to it. The electric vehicle is used to track the energy consumption for each route. The total energy consumption was then averaged over the number of routes and thus generating an average energy consumption for the entire map. In table 3.1, the length of the random routes is presented.
3.4.5 Comparison & Analysis
The comparison model was used to compare the results over the two comparison criteria; (1) energy consumption and (2) ease of access. The results of (1) were plotted for each parameter and later analysed and discussed in the context of validity and ethics. The results of (2) were analysed by how easy it is to acquire the parameter.
3.5 Validity Criteria
Within empirical research, there are certain tests which are used to establish the quality of the research. Within quantitative research approach, these criteria are:
internal validity, external validity, dependability, confirmability and construct validity[24]. These criteria correspond to criteria in qualitative research.
• Credibility corresponding to Internal validity
• Transferability corresponding to External validity
• Reliability corresponding to Dependability
• Objectivity corresponding to Confirmability
Validity within our research corresponds to the correctness of how our guidelines mirror the real-world characteristics of an electric vehicle. Since our study will be qualitative, we will use the corresponding qualitative criteria.
3.6 Ethics & Sustainability
The ethical guidelines followed by us during our research were lifted from the Swedish Research Council in their publication Good Research Practice[25]. The book gives examples and guidelines on how a researcher can maintain ethical
considerations throughout their research. The book can be summarised into these guidelines:
• You shall tell the truth about your research.
• You shall consciously review and report the basic premises of your studies.
• You shall openly account for your methods and results.
• You shall openly account for your commercial interests and other associations.
• You shall not make unauthorised use of the research results of others.
• You shall keep your research organised, for example through documentation and filing.
• You shall strive to conduct your research without doing harm to people, animals or the environment.
• You shall be fair in your judgement of others’ research.
4 Research Process
In this chapter, the data that was collected during the Data Collection phase is presented. This data influenced the guidelines that are presented in Chapter 5.
In Section 4.1 the parameters that are considered in the guidelines are defined and explained. In Section 4.2 the results of the simulations of the parameters are presented.
4.1 Parameter Definitions
In this section, the parameters that were studied are defined. The parameters chosen are used because they are implemented in SUMO in an easily accessible way. When modelling a vehicle to simulate, these parameters can be specified by the end-user. This means no change of source code is needed to alter the parameter values.
The parameters air drag coefficient, constant power consumption, front surface area, internal moment of inertia, propulsion efficiency, radial drag coefficient, recuperation efficiency, roll drag coefficient, vehicle mass are listed in Table 4.1 and their characteristics are described in Sections 4.1.1-4.1.9.
Table 4.1: The parameters that are used in the guidelines Parameter Variable Air drag coefficient Cw Constant power consumption Pconst
Front surface area A Moment of inertia of internal elements Jint
Radial drag coefficient Crad Recuperation efficiency ηprop Propulsion efficiency ηprop Roll drag coefficient Croll
Vehicle mass m
4.1.1 Air Drag Coefficient
The Air Drag Coefficient is used in calculations of the air drag for a vehicle. The number is used to model all of the complex dependencies of shape, inclination and flow conditions on vehicles [26].
D = Cw ρ V2
2 A (1)
The Drag Equation (1), is used to calculate the drag D of an object. Cw corresponds to the drag coefficient, ρ to the air density, V to the velocity and A to the front surface area. The force (and energy) needed to overcome this is therefore equal to D.
The value of the parameter is most often found by experimentation since the actual calculations are very complex. For passenger vehicles, the value of the Air Drag Coefficient is usually between 0.22 to 0.45 [27].
4.1.2 Constant Power Consumption
The Constant Power Consumption parameter is a constant that represents the power consumption of all the auxiliary systems of the vehicle. This includes the air conditioner, electronics, spotlights, and so forth [28]. In real life, the power consumption of auxiliary systems is not constant. The energy consumption changes depending on a number of variables like temperature and visibility conditions. The energy consumption of auxiliary systems varies between models for vehicles as well. So a range of values for the Constant Power Consumption is hard to find.
4.1.3 Front Surface Area
The Front Surface Area corresponds to the area of an object that experiences drag. In a vehicle, this corresponds to the are visible from the front of the vehicle [29]. The Front Surface Area is used to calculate the drag of a vehicle using the Drag Equation (1). The value of the parameter is most often measured by the manufacturer, and is usually 1-3m2 depending on the model [29].
4.1.4 Internal Moment of Inertia
The Internal Moment of Inertia is a measurement of the force required to turn an object. For a simple pendulum, the moment of inertia I is calculated using the mass the pendulum m times the length of the pendulum r squared (2).
I = m r2 (2)
In a more complex object such as a vehicle, the moment of inertia is calculated by performing this equation on all of the components of the vehicle and adding the values. The length of the pendulum now refers o the distance of the component to the centre of mass of the vehicle and the mass of the pendulum refers to the mass of the component [30]. This quickly becomes a very time consuming and complex
process, and as such, the moment of inertia is most often measured or estimated instead [31]. The range of Internal Moment of Inertia is generally between 500 to 5000 kg m2 [30].
4.1.5 Propulsion Efficiency
Propulsion Efficiency is a measurement of how efficiently a vehicle can convert the energy stored in the fuel to movement energy. For an electric vehicle, the efficiency is dependant on how much energy is lost to different components of the vehicle. For example, when charging an electric vehicle, around 10% of the energy is lost and in the engine of the vehicle, around 20% of the energy is lost.
The Propulsion Efficiency of an electric vehicle is between 60 to 73% depending on the driving conditions [32].
4.1.6 Radial Drag Coefficient
The Radial Drag Coefficient is a constant used to determine the friction in the wheels when the vehicle is turning. The formula used to calculate the loss in energy from radial drag (3) is based o the formula for centripetal force Fc [33].
Fc = Cradmv2
r (3)
In the formula (3), Cradrepresents the Radial Drag Coefficient, m represents the weight of the vehicle, v represents the angular velocity and r represents the radius of the turn. The phenomena comes from the wheels of the vehicle experiencing a sideways force when turning and therefore the bearings in the wheel will roll less smoothly.
4.1.7 Recuperation Efficiency
The Recuperation Efficiency is a constant that corresponds to how efficiently the vehicle can regain energy when slowing down. In electric vehicles, this is achieved using a generator that taps into the wheels of the car and drawing energy from them when the car is slowing down or rolling down a slope [34]. The efficiency of this process is usually between 60 to 70% [35] [36].
4.1.8 Roll Drag Coefficient
The Roll Drag Coefficient is a measurement on how much energy is lost by the wheels rolling on a surface. The Roll Drag Coefficient is dependant on the
resistance in the bearings, the rolling surface and the weight of the vehicle.
Fr = Crollm ag (4)
The Roll Drag Coefficient is used to calculate the rolling resistance Fr using the formula above (4). In the formula (4), Crollrepresents the Roll Drag Coefficient, m represents the weight of the vehicle and ag represents the gravitational acceleration on the vehicle. Depending on the surface, the Roll Drag Coefficient can vary, but for a car on concrete the value of coefficient is 0.01 to 0.03 [37].
4.1.9 Vehicle Mass
The Vehicle Mass is a vital part in many of the formulas that calculate the energy consumption in the other parameters. Apart from these formulas, the inertia of a vehicle is largely dependant on the weight. Overall the weight determines how much force is required to accelerate the vehicle in any direction or rotation. The weight of a personal vehicle is around 800 to 2500 kg [38].
4.2 Simulation Results
When creating the simulation, a vehicle needed to be modelled. Since we used a Tesla Model S to drive the real car and verify SUMO we decided to reuse those parameter values, as described by Table 4.2 as starting values for the simulation.
Table 4.2: Starting parameter values. Target model: Tesla Model S 70D Parameter Variable Value (unit) Source Air drag coefficient Cw 0.24 [29]
Constant power consumption Pconst 50 (W s) 50% Default [20]
Front surface area A 2.34(m2) [29]
Moment of inertia of internal elements Jint 0.01 (k∗ gm2) Default [20]
Radial drag coefficient Crad 0.5 Default [20]
Roll drag coefficient Croll 0.01 [37]
Propulsion efficiency ηprop 0.80 [32]1 Recuperation efficiency ηrecoup 0.64 [34]
Vehicle mass m 2240 (kg) [38]2
In Figure 4.1, the results from the real-life electric vehicle test are presented with the simulation results from Simulation of Urban Mobility (SUMO). The
2We added 140kg because two adult people in the car
2This is slightly higher than the ranges described due to Tesla claiming to be more efficient
routes driven were of different range and therefore also have different energy consumption. Below are graphs and tables of the raw results of our simulated tests for each parameter.
0 500 1,000 1,500
Täby-10.8km Bergshamra-5.2km Djursholm-5.8km
1,787.97 760.99
905.24
1,700 1,100
800
Energy consumption (Wh)
Simulation Real car
Figure 4.1: Simulation versus real world by destination
To simulate several parameters with different ranges at the same time, a python script was created to automate this operation. The pseudo-code for this can be seen in Algorithm 1.
Data: Sumo configuration file Result: 2D array
1 FunctionParameterImpact(Filename of SUMO configuration)
2 parameterlist= (parameter, minValue, maxValue, n);
3 for each parameter in parameterlist do
4 values= n linearly spaced points between minValue and maxValue;
5 for each value values do
6 Run SUMO simulation
7 Calculate and average cost over all routes
8 Store value and cost in matrix
9 end
10 Output result
11 end
12 end
Algorithm 1: Simulation Script
0.2 0.25 0.3 0.35 0.4 0.45 1,150
1,200 1,250 1,300
Air Drag Coefficient
Watt-hours
0 50 100 150
1,160 1,170 1,180
Constant Power Consumption
Watt-hours
1 1.5 2 2.5 3
1,050 1,100 1,150 1,200
Front Surface Area
Watt-hours
1,000 2,000 3,000 4,000 5,000 1,500
2,000 2,500 3,000
Internal Moment of Inertia
Watt-hours
0.6 0.65 0.7
1,300 1,400 1,500 1,600
Propulsion Efficiency
Watt-hours
0.2 0.4 0.6 0.8 1
1,100 1,200 1,300
Radial Drag Coefficient
Watt-hours
0.6 0.62 0.64 0.66 0.68 0.7 1,140
1,160 1,180
Recuperation Efficiency
Watt-hours
1 1.5 2 2.5 3
·10−2 1,500
2,000
Roll Drag Coefficient
Watt-hours
1,000 1,500 2,000 2,500 600
800 1,000 1,200
Vehicle Mass
Watt-hours
The two axes show how much energy the vehicles consumed with the current parameter values. The x-axis of the graphs represents the value of the parameter over a range of possible values for electric vehicles. The ranges are shown in Table 5.1. Each point of the y-axis represents the average energy consumption for 50 trips (mentioned in Section 3.4.4) for the current simulation iteration.
These graphs show that most of the parameters have a linear growth. The exceptions to this are Propulsion Efficiency, which is visibly non-linear in the graph.
5 Result, Analysis & Discussion
In this chapter the results of the thesis are presented, analysed and discussed. The results from the study are presented in Section 5.1. The results are analysed based on the comparison criteria in the comparison model in Section 5.2. Lastly, the results are discussed in Section 5.3.
5.1 Results
In this section, the results on the simulations are summarised into two tables 5.1 5.2. These tables are the grounds from where the guidelines are determined.
Table 5.1: Slopes and intercept
Parameter Slope value (non-linear) Intercept Roll drag coefficient 46351.81 703.61 Propulsion efficiency (-1870.86) 2672.65
Air drag coefficient 762.15 984.42
Recuperation efficiency -490.4 1481.2
Radial drag coefficient 234.45 1050.21
Front surface area 78.12 984.54
Vehicle mass 0.44 191.68
Moment of inertia of internal elements 0.35 1166.88
Constant power consumption 0.17 1158.63
The slopes of the parameters, including the non-linear ones, were calculated using linear regression. The slope and intercept of all parameters are shown in Table 5.1. These values show that there is great variety in the characteristics of the parameters, even if most of them are linear.
Table 5.2: Difference in energy consumption over parameter range Parameter Energy (Wh)
Moment of inertia of internal elements 1595.30 Roll drag coefficient 927.00
Vehicle mass 740.44 Propulsion efficiency 351.70 Radial drag coefficient 210.99 Air drag coefficient 175.30 Front surface area 156.25 Recuperation efficiency 49.04 Constant power consumption 25.29
In Table 5.2 the total change in energy consumption for every parameter is displayed. This is calculated using the found ranges. The results are ordered from lowest energy consumption to the highest. By these results, the parameter which impacts the energy consumption most is the internal moment of inertia, which has a difference of 1595.30Wh over the range of values. The parameter with the least difference in energy consumption, with a difference of 25.29Wh.
5.2 Analysis
In this section, the results are analysed. The analysis is split into three sections each corresponding to the analysis of different subjects. In Section 5.2.1 the results from the real electric vehicle test is analysed. The other two sections correspond to the comparison criteria from the comparison model. In Section 5.2.2 the results of the simulations in Simulation of Urban Mobility(SUMO) is analysed in the context of relative impact on the energy consumption. In Section 5.2.3 the parameters from the SUMO simulation are analysed in the context of how difficult the parameter values are to acquire for third party developers of an electric vehicle route planner.
5.2.1 Real Electric Vehicle Test Results
The results from the real electric vehicles test show that there is a correlation between the simulation an real life. The anomaly among these results is the route driven in Bergshamra, where the simulation produced an energy consumption that is 339.01Wh ( 31%) less than the real-life electric vehicle. The other two trips showed closer results where SUMO presented results that were around 5- 12% higher than the real test route. This means that there is a slight correlation between the results, but that the values can be very different from each other on different routes.
5.2.2 Relative Impact on Energy Consumption
The results of the study points towards that most of the vehicle’s parameters have linear growth, with the exceptions having almost linear growth. The results also show that there is a clear difference in the relative impact on the energy consumption. This means that some of the parameters are less valuable to the overall energy consumption than others. For example, the vehicle mass has an energy difference that is 756% larger than the radial drag coefficient. From the perspective of a route planner, the vehicle mass will be around 7 times more important to implement than the radial drag coefficient.
Table 5.3: Impact on energy consumption relative to mean energy difference (470.15Wh)
Parameter Relative Impact Moment of inertia of internal elements 3.40
Roll drag coefficient 1.97 Vehicle mass 1.57 Propulsion Efficiency 0.75 Radial drag coefficient 0.45 Air drag coefficient 0.37 Front surface area 0.33 Recuperation efficiency 0.10 Constant power consumption 0.05
By this logic, it is possible to estimate the relative importance of all parameters that are considered in this study. Visualised in Table 5.3, the impact of some variables are almost insignificant to the entire energy consumption. However, these parameters still need to be considered in an accurate prediction.
5.2.3 Finding Parameter Values
The difficulty to access a parameter value can have an impact on its overall worth. If the parameter is easily accessible but have no impact on the energy consumption, it might not be worth implementing. If the parameter is very difficult to acquire but have huge implications on the result it might be worth the time. The accessibility can, therefore, be crucial to the value of the parameter.
Below is an explanation for how we did (or did not) find the values for the parameters followed by a summary in Table 5.4.
Vehicle mass is always provided by the manufacturer and therefore a very easy parameter to acquire.
Front surface area is the largest horizontal view of a car [29]. This parameter is often listed, either by the manufacturer [39] or by someone else studying the aerodynamics of cars [29]. Since most cars have been tested in wind tunnels this parameter should be official. If it is not, however, anyone can take a picture from the front of the vehicle and calculate the area from that image. This makes it very easy to acquire.
Air drag coefficient is hard to calculate for the average person but is rather easily gathered from other sources. The result can, however, have some variations
in the results, depending on which wind tunnel and how the test and analysis is done [40]. This coefficient is very sought out by manufacturers since they want the car to be as aerodynamic as possible and will probably brag if it is low (good).
Wikipedia also has a list of numerous cars coefficients [27].
Internal moment of inertia (Jint) is virtually impossible to find out. Jint is calculated summing up all internal elements within a car multiplied with the objects relative (square) distance to the centre [30]. This makes it extremely difficult and there is only one source, from 1998, that we found to list such values [41]. There are also four different types Jint listed in the paper; pitch, roll yaw and ”Roll/Yaw Product” which has a wide range and SUMO does not provide any information to which value they use. Tesla claims they have ”extremely low polar moment of inertia” in the Model 3, but does not provide any actual value [42]. In conclusion, to find out the Jint would require extensive research and calculations.
Radial drag coefficient is a parameter value we could not find any information about. We found it is based on the centripetal force by studying the source code of SUMO and comparing the implementation with a formula for the centripetal force, Fc= mvr2 [33]. This did not make it possible to find a value suitable for any specific car however since it is dependant on speed and radius, both of which are variables and not constants. Radial drag coefficient is therefore extremely hard to acquire.
Roll drag coefficient is hard to define exactly but an approximation is easy to find. There are some typical values for car tires depending on the surface [37].
However, there seems to be no official information about any specific car model and what tires they usually use that can narrow the range.
Constant power consumption is hard to find an exact value for since it depends on what functions are used in the car. A/C and other non-engine consumers will increase it for example. We found no information about what this number could be for any car.
An exact number for recuperation efficiency is hard to find. The closest to the Tesla Model S was a Tesla blog post on the Tesla Roadster [34]. The BMW i3, which is a vastly different car, have a similar recuperation efficiency so it seems reasonable that more cars are within the same span [35]. Official numbers can be hard to find but it at least seems tests are being performed to find out.
Propulsion efficiency is the same story as recuperation efficiency. There is also more information about electric motors in general rather than specific cars
although some test have been performed [32]. Efficiency in terms of Wh/km/kg has also been studied and it may be easier to access [43].
Table 5.4: Parameter by difficulty to acquire Parameter Difficulty to find Vehicle mass Very easy
Front surface area Easy Air drag coefficient Easy
Roll drag coefficient Hard but approximate is easy Constant power consumption Hard
Recuperation efficiency Hard Propulsion Efficiency Hard Radial drag coefficient Very hard Moment of inertia of internal elements Very hard
5.3 Discussion
In this section, the results of the study are discussed. The discussion is split into three sections. In section 5.3.1 the results from the real electric vehicle test are discussed. The other two sections correspond to the comparison criteria in the comparison model. In section 5.3.2 the relative impact of the parameters is discussed. In section 5.3.3 the relative difficulty in finding the parameter values is discussed.
5.3.1 Real Electric Vehicle Test
There have been documented issues with the accuracy of the energy consumption for electric vehicle simulation in SUMO [17]. The trend that was is that SUMO consistently predicted 23-28% lower energy consumption than a real-life vehicle.
We wanted to test for our selves and we got access to a Tesla Model S to do so.
Our results show that the paper mentioned previously was correct to some extent.
A big difference between the two first trips and the third and final one were noted, see Figure 4.1. The last route, Bergshamra, is the only case in our simulation where the energy consumption is lower than the real vehicle. The distance travelled in both the Djursholm and Bergshamra routes is almost the same, but the energy consumption is drastically larger in the Bergshamra route even though it is the shortest distance. This was the last route driven and we acknowledged that the A/C started at the beginning of this trip but did not turn it off. This could explain why the real energy consumption is higher than the simulated one, SUMO does not account for the A/C accurately.
As stated in the delimitations, weather conditions were not considered when performing this test. But it should be stated that outside weather conditions were close to the SUMO implementation when the test was performed. If the test would have been performed during the winter, another result might have occurred, due to temperature and road conditions.
5.3.2 Relative Impact on Energy Consumption
The results show that some of the parameters have more influence on the energy consumption than others. The reliability of these results are based on the accuracy of the simulation. As previously discussed, the accuracy of SUMO might be less than optimal. Although we found a correlation between the SUMO simulation and the real vehicle, there is still a measurable difference between the two. This means that based on the real electric vehicle test results, some of the parameters might have less impact on the energy consumption. This is because, in the real vehicle test, the results insinuate that SUMO is producing higher energy consumption in the simulation.
The growth of the parameters are as expected from a physics standpoint [44].
For example, the roll drag coefficient is expected to have a very large slope since the physically possible values for the coefficient is 0 to 1, which also applies to propulsion efficiency.
5.3.3 Finding Parameter Values
The ranges of the parameters that were chosen are based on sources of vehicle statistics. However, these ranges are only approximations and could have outliers that don’t fit our model, or the range could be smaller. Preferably, the ranges would be based on statistical distributions. That way, more accurate estimations can be made, and statistical outliers can be predicted. However, such sources are very hard to find and often very old. So our estimated ranges for the parameters are likely to be sub-optimal. However, the slope of the parameters will still be the same for most cases since most of the parameter’s growth is linear. Our estimation of relative impact should hold up as long as the ranges are not changed significantly.
Since some parameters are harder to find but affect the energy consumption in a big way, we decide if the trade off is worth it or not. In Table 5.5 we give our recommendations on how to approach each parameter in a route planner.
Table 5.5: Guidelines for implementing electric vehicle parameters in a routing application
Parameter Implementation Action
Vehicle mass Use as variable Always implement Front surface area D = CwρV22A Always implement Air drag coefficient D = CwρV22A Always implement
Roll drag coefficient F = CrollN Approximate over lane based on the road surface
Constant power
consumption
Implement a better function than just constant [17]
Recuperation efficiency
Eout = Ein∗ ηrecoup Approximate on general level as best possible
Propulsion Efficiency Eout = Eout∗ ηprop Approximate on general level as best possible
Radial drag coefficient Fc = Cradmvr2 Ignore Moment of inertia of
internal elements
Ignore until more research exist
If the manufacturer’s released more information about the vehicles, the implementation of a route planner would be much easier. Instead of trying to read on different sites, calculate by yourself or just try to approximate a value, the manufacturers could provide more information that would help third party route planners to make better implementations of each variable.
6 Conclusions and Future Work
In this chapter, the conclusions of the thesis are presented in Section 6.1 and examples of future work is described in Section 6.2.
6.1 Conclusion
There is an urgent need to migrate the vehicle industry to electrical vehicles due to climate changes associated with internal combustion engines. Customers with internal combustion engines or no car at all are however discouraged from acquiring an electric vehicle since they don’t think they will be able to go on long trips due to battery capacity.
The third-party tools or applications that provide trip planning do not provide customers with confidence enough to make them feel good about going on a long trip.
It is evaluated how much different parameters for electric vehicles affect the energy consumption of an electric vehicle within SUMO, an open-source traffic simulator.
The thesis shows that the parameters, although many affect the energy consumption vastly, are often hard to come by. Specifically when looking for exact values.
Basic guidelines are provided. Guidelines that researchers can develop further with more parameters and more accurate value ranges. These can help developers quickly understand what parameters they need to implement in their algorithm if they want an accurate estimation of electric vehicles for a route planner.
6.2 Future Work
The hope is that the manufacturer’s release, or research if they do not have the data immediately, more data of their vehicles that could be used to effectively implement a route planner for electric vehicles.
To be more reliable as a simulation tool, SUMO should improve their energy consumption calculations. This would help the type of research done in this thesis.
With a more accurate simulation program, real-world problems could be more easily explored by the average developer.
Deeper research in what parameters exist and what range they have. The ranges in this thesis are not an exact representation of the real world since they were not
found. Internal moment of inertia, which seem to have a very large range of values according to the paper from 1998 [41], and could, therefore, have a big impact on the electrical vehicle route planners with new research.
The guidelines need to be further evaluated because of SUMO’s limitations and error margin. It is also encouraged that researchers extend the guidelines into a framework. One extension that we can see first and foremost is with particle swarm optimisation [11, 12].
Weather conditions is another aspect that is of interest. Cold, warm or cool temperature together with different humidity could possibly affect the total energy consumption and make for a whole new layer of implementation. Researching how to implement different weather conditions correctly would therefore be very interesting.
References
[1] Ashkrof, Peyman, Gonçalo Homem de Almeida Correia, and
Bart van Arem. “Analysis of the effect of charging needs on battery electric vehicle drivers’ route choice behaviour: A case study in the Netherlands”.
In: Transportation Research Part D: Transport and Environment 78 (Jan. 2020).
[2] 5 anledningar till att jag inte kommer att köpa elbil. URL:https:
//m3.idg.se/2.1022/1.702617/anledningar-att-inte-kopa-elbil.
[3] Lin, Zhenhong. Rethinking FCV/BEV Vehicle Range: A Consumer Value Trade-off Perspective. 2010.
[4] Bedogni, Luca et al. “Driving without anxiety: A route planner service with range prediction for the electric vehicles”. In: IEEE, Nov. 2014,
pp. 199–206.
[5] Faraj, Mahmoud and Otman Basir. “Range anxiety reduction in battery-powered vehicles”. In: IEEE, June 2016, pp. 1–6.
[6] Tesla Model S. URL:https://www.tesla.com/models.
[7] Franke, Thomas, Nadine Rauh, and Josef F. Krems. “Individual
differences in BEV drivers’ range stress during first encounter of a critical range situation”. In: Applied Ergonomics 57 (Nov. 1, 2016), pp. 28–35.
[8] Rauh, Nadine, Thomas Franke, and Josef F. Krems. “Understanding the Impact of Electric Vehicle Driving Experience on Range Anxiety”. In:
Human Factors 57.1 (Feb. 1, 2015), pp. 177–187.
[9] How to Build an EV Trip Planner System That Users Will Fall in Love With. Sept. 26, 2019. URL:https://www.intellias.com/how-to-build- an-ev-trip-planner-system-that-users-will-fall-in-love-with/.
[10] Delling, Daniel et al. “Customizable Route Planning”. In: ed. by
Panos M. Pardalos and Steffen Rebennack. Springer Berlin Heidelberg, 2011. ISBN: 978-3-642-20661-0 978-3-642-20662-7.
[11] Rawashdeh, Osamah and Rami Abousleiman. “Electric vehicle modelling and energy-efficient routing using particle swarm optimisation”. In: IET 10 (Mar. 1, 2016).
[12] Garcia, Aries G., Lew Andrew R. Tria, and Marc Caesar R. Talampas.
“Development of an Energy-Efficient Routing Algorithm for Electric Vehicles”. In: IEEE, June 2019, pp. 1–5.
